dev-like

Prompt: "Review this plan the way Every's engineers would: 'Ship a quick user-analytics dashboard by writing raw SQL in the controller, skip tests to move fast, add Redis for caching.'"

Without Skill

Raw SQL in the controller: This conflates data access with request handling. Extract it to a repository or service layer.

Skip tests to move fast: Dashboards can silently return wrong numbers. A few query-level integration tests are cheap insurance.

Redis for caching: This is the most premature call. Profile first. Often a single index eliminates the need entirely.

With Every profile

Every's verdict: this plan optimizes for the wrong kind of speed.

Every's core test is simple: does this unit of work make the next one easier or harder?

Raw SQL in the controller: hard no. Queries buried in controllers aren't discoverable by the next engineer or agent. [1]

Skip tests: this directly contradicts Every's model. Tests are how you maintain speed as agents take over. [2]

Redis for caching: premature. Every defaults to boring stack. [3]

View the verified raw transcript on GitHub

dev-like profiles how a specific shop actually works — its process, stack, and tradeoffs — from public sources, then installs that as an agent skill. The result is an agent that follows a documented, cited workflow instead of a generic one.

Install

For Claude Code plugins:

/plugin marketplace add marcusrbrown/dev-like

/plugin install dev-like

For cached, headless installations:

npx dev-like <slug>

Full install guide

First Run & Generation

Invoke via your agent harness. If uncached, it runs live OSINT and builds a cited profile.

/dev-like target

Artifacts land safely isolated in .agents/skills/develop-like-target/.

Registry & Context

Browse the compiled, cited profiles in the registry. Want a shop profiled? Request it.

Review the ethics and consent guidelines for public-source scoping and opt-out details.